Evaluation of accuracy of as-built 3D modeling from photos taken by handheld digital cameras

Abstract

As-built documentation is an essential set of records, consisting of construction drawings, specifications and equipment location, which are kept for facility management purposes. These documents are constantly being created and modified throughout the life of a project. This process is usually manual and fraught with errors, which inhibits reliable decision making. Technological advancements have made it possible to generate 3D models to assess as-built conditions for construction monitoring purposes, such as verifying conformance to baseline project schedules and contract specifications. For this purpose, 3D point clouds are widely generated using laser scanners. However, this approach has limitations in the construction industry due to expensive and fragile equipments, lack of portability and need of trained operators. This study aims at investigating an alternate technology to generate as-built 3D point clouds using photos taken using handheld digital cameras, compare them against the original as-built 3D models, and check for accuracy of the modeling process. This analysis can aid in a more reliable and effective decision making due to its cost effectiveness and ease of use, particularly in heavy infrastructure projects which are continually undergoing rehabilitation work. To achieve these objectives, a set of guidelines are developed for taking photographs that enable effective generation of 3D point clouds using off-the-shelf software packages. The accuracy of the modeling process is investigated using the results of the as-built 3D point cloud modeling of a 2000 ft under construction bridge in southern United States. Finally, the range of tolerance and deviation of element dimensions is determined by comparing the photo based model to the actual as-built model (developed using 2D drawings). Furthermore, to compare point clouds generated from laser scanning and photogrammetry, a laser scan and an image based survey of an exterior wall of a university building were also done. Results show that this technology in its present state is not suitable for modeling infrastructure projects, however technological developments can enable this to be an efficient way to extract measurements of inaccessible objects for progress monitoring purposes and the models can also be stored for future dimension takeoffs for decision making and asset management purposes.